Martin Sebor via Gcc-patches <gcc-patches@gcc.gnu.org> wrote:
On 11/3/20 4:13 AM, Richard Biener wrote:
This fixes the bad assumption that sizeof (bool) == 1
Bootstrap / regtest running on x86_64-unknown-linux-gnu.
2020-11-03 Richard Biener <rguent...@suse.de>
PR bootstrap/97666
* tree-vect-slp.c (vect_build_slp_tree_2): Scale
allocation of skip_args by sizeof (bool).
---
gcc/tree-vect-slp.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/gcc/tree-vect-slp.c b/gcc/tree-vect-slp.c
index e97fbe897a7..08018a1d799 100644
--- a/gcc/tree-vect-slp.c
+++ b/gcc/tree-vect-slp.c
@@ -1428,7 +1428,7 @@ vect_build_slp_tree_2 (vec_info *vinfo, slp_tree
node,
/* If the SLP node is a PHI (induction or reduction), terminate
the recursion. */
- bool *skip_args = XALLOCAVEC (bool, nops);
+ bool *skip_args = XALLOCAVEC (bool, sizeof (bool) * nops);
This change caught my eye. XALLOCAVEC() does the scaling:
#define XALLOCAVEC(T, N) ((T *) alloca (sizeof (T) * (N)))
so scaling it again doesn't make sense to me...
memset (skip_args, 0, nops);
... especially not when it's followed by the memset call above that
doesn't do the same. If I'm missing some subtlety it might be worth
adding a comment explaining it.
um yes, the fix I tested on powerpc-darwin was…
memset (skip_args, 0, sizeof (bool) * nops);
(I haven’t tried current master, since you pushed this change - regstrap
takes most
of a day on that machine).
Iain
